Download - All-optical devices in polymer optical fibernathan.instras.com/documentDB/paper-363.pdf · We report on polymer optical fiber devices for sensors, optical switchesrlogic, and optical

Ž .Chemical Physics 245 1999 533–544www.elsevier.nlrlocaterchemphys

All-optical devices in polymer optical fiber

Mark G. Kuzyk a,b,), Dennis W. Garvey a, Steven R. Vigil a, David J. Welker c,1

a Department of Physics, Washington State UniÕersity, Pullman, WA 99164-2814, USAb Materials Science Program, Washington State UniÕersity, Pullman, WA 99164-2814, USA

c Sentel Technologies L.L.C., Box 41-1, 1415 Eastgate BlÕd, Pullman, WA 99163, USA

Received 15 November 1998

Abstract

We report on polymer optical fiber devices for sensors, optical switchesrlogic, and optical actuators. In this paper, wegive a brief overview of polymer fibers, discuss recent all-optical switching results, and describe how an optical sensor andactuator can be built into a single fiber device. Future technologies that are made possible with such optical devices andphotomechanical mechanisms are also discussed. q 1999 Elsevier Science B.V. All rights reserved.

1. Introduction

In another article in this issue, we discuss how allelectronic devices can be subdivided into the fourclasses: Data transmission and interconnects, sen-

w xsors, computingrlogic, and actuation devices 1 . Inthat article, an overview of the use of polymer fibersfor interconnects and transmission was presented andseveral linear-optical characterization experimentalresults were reported. The first demonstration ofelectro-optic modulation in a polymer fiber withelectrodes was also presented.

In this paper, we present all-optical switchingresults, and demonstrate how an optical fiber can beused as a smart actuator, in which an actuator andsensor are built into the same fiber. The results ofthis paper, together with the other article in thisissue, show that all four device classes have beendemonstrated in polymer fiber.

) Corresponding author. E-mail: [email protected] http:rrwww.senteltech.com

This paper is organized into two major sections.The first deals with optical logic and switching andthe second deals with photomechanical effects, sen-sors, and actuators.

In the optical logic and switching section, wereport on Sagnac experiments used to characterizethe intensity-dependent refractive index of a single-mode dye-doped fiber. Subsequently, all-opticalswitching in a similar Sagnac interferometer isdemonstrated. Finally, the first demonstration of non-linear switching between two parallel cores in adual-core polymer fiber is reported.

The photomechanical-effect section describes themechanisms that lead to a photomechanical response.Subsequently, actuator-based devices that use thephotomechanical effect are described and data pre-sented.

2. Optical logic and switching

In this section, we report on progress in opticalswitching studies in organic materials. We first dis-

0301-0104r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0301-0104 99 00058-0

( )M.G. Kuzyk et al.rChemical Physics 245 1999 533–544534

cuss all-optical switching in bulk media followed byprogress in switching studies in fibers.

2.1. All-optical switching

ŽThe Sagnac interferometer sometimes called a.loop mirror has been shown to be a highly efficient

geometry for optical switching because it is immuneto external influences such as temperature variationsand mechanical vibrations. Many examples of suchswitches made in silica-based fiber can be found in

w xthe literature 2–4 . It has also been used as a tool tomeasure the third-order susceptibility of a dye-doped

w xwaveguide 5 . In separate measurements, we haveshown that the third-order susceptibility of dye-dopedpolymer optical fiber is, within an order of magni-

w xtude, comparable to liquid nitrobenzene 6 . Wetherefore begin by showing optical switching in bulksamples as a proof of principle.

The top portion of Fig. 1 shows a diagram of theall-optical switching experiment. A light pulse islaunched into an asymmetric Sagnac interferometer.A small portion of the incident beam is removedfrom the Nd:YAG laser with a glass slide and moni-tored with a detector. The counterclockwise pulsegoes through an attenuator before entering the sam-ple so that the pulse intensity is too small to induce aphase shift due to the intensity-dependent refractiveindex. The clockwise pulse, however, experiences anintensity-dependent phase shift due to the sample.After this pulse leaves the sample, it also passesthrough the attenuator so that both pulses are thesame intensity when they arrive back to the beamsplitter — at which point they interfere. The amountof light reaching the output detector depends on thephase difference between the two pulses. For aninterferometer with no sample, both pulses construc-tively interfere and all the light exits to the outputdetector. If the two pulses totally destructively inter-fere, all the intensity will go back to the laser source.Note that the weak pulse passes through the samplefirst. If the time ordering were reversed, the weakpulse would see the transient effects of the strongpulse. Indeed, this other arrangement could be usedto measure the dynamical response of the sample.With the present set-up, only the instantaneous re-sponse within the 25 ps pulse duration is measured.

Ž .Fig. 1. All-optical switching experimental diagram top and dataŽ .bottom .

For a Kerr-material sample, the phase shift islinear in the incident intensity so the intensity at theoutput detector should be a sinusoidal function of theincident intensity. The bottom portion of Fig. 1shows the ratio of the intensity measured by theoutput detector to the intensity measured by the inputdetector for a liquid nitrobenzene sample in a quartzcell. The theoretical curve is a sinusoidal function.Over an intensity range of about 300 MWrcm2, theintensity dependent phase shift in the 2 cm longliquid cell is about pr2 radians. Given that polymeroptical fibers have a comparable third-order suscepti-bility, a 20 cm fiber should yield a phase shift thatcorresponds to full switching. Below, we summarizeresults of single-mode fiber measurements and dis-cuss how the state-of-the-art fiber materials fareagainst the nitrobenzene results.

2.2. Fiber Sagnac measurements

The Sagnac interferometer provides the idealmeasurement scheme for single-mode fibers. Not

( )M.G. Kuzyk et al.rChemical Physics 245 1999 533–544 535

only is the technique sensitive enough to measuresmall nonlinearities, but it is also similar to the finaldevice geometry. Fig. 2 shows a schematic view ofthe experiment.

The mode-locked pulse train from the laser sourceis passed through a rotating half-wave plate andpolarizer to sinusoidally modulate the pulse train.The polarizer is set so that the electric field of thelight is 458 with respect to the table surface. Apolarizing beam splitter separates the two polariza-tions. The ratio of intensities of the light passingthrough the polarizer is set to 1r2–1r2 by adjusting

Ž .the half-wave plate not shown . The two counter-propagating waves travel through several compo-

Ž .nents described later and are recombined at thebeam splitter. The polarizer in front of the detector

combines the two polarizations so that they caninterfere. When the light intensity is low and thephase difference between the two beams is zero,constructive interference results and all of the lightgoes to the detector.

The inset shows the sinusoidal envelope of theinput beam pulse train. Because the repetition rate of

Ž .the laser is high 76 MHz and the signal exiting thedetector is integrated with a time constant that islong compared with the pulse spacing but shortcompared with the period of the sinusoidal envelope,we can view the laser source as continuous with asinusoidal modulation in time. As in the all-opticalswitching experiment, an attenuator makes the inter-ferometer asymmetric. Because the counterpropagat-ing pulses are of orthogonal polarization, their rela-

Fig. 2. Sagnac–Kerr measurement of single-mode fibers.


tive phase can be adjusted by a pair of birefringentŽ .wedges the compensator in the figure .

The experiment proceeds as follows. First, theamplitude of the detector output is measured with alock-in amplifier that is synchronized to the fre-quency of the incoming sinusoidal modulation pro-vided by the rotating half-wave plate. The phase shiftis varied during this measurement to produce an

interferogram. The amplitude of the interferogram isa measure of the input intensity in arbitrary unitsŽ .usually measured in detector voltage . The top partof Fig. 3 shows typical data and a theoretical fit. Thelock-in then measures signal at twice the modulationfrequency as a function of the wedge position. Themiddle plot in Fig. 3 shows typical data and theory.The double peaked function is used to get both the

Ž . Ž .Fig. 3. Interferogram at the modulation frequency, twice the modulation frequency, and the phase-theory curve and data points .


real and imaginary parts of the third-order suscepti-bility. The bottom plot shows the measured phaseand the theory for the phase. The important featureto note is that the peaks in the middle plot shouldapproximately align with the maximum slope of thesinusoidal upper curve and the cusps in the middlecurve should coincide with a 1808 phase change. Thedata presented in all three curves are thus self-con-sistent.

The details of the data analysis are presentedw xelsewhere 6 . Here, we tabulate some new results

from this measurement technique on several single-mode fibers. Fig. 4 shows four dyes along with thephase shift measured for a 20 cm fiber with 10 mmcore diameter and about 0.1 wt.% dye doping level.Note that the phase shift is expressed as a percentageof a p phase shift required for full switching. Thepeak power of the ls1.064 mm laser is 100 mWand for ISQ fiber yields a switching efficiency of50%rW.

While these switching efficiencies are promising,there are other issues that need to be considered toevaluate squaraine dye-doped fibers for their useful-ness in optical switching devices — such as two-

Ž .photon absorption TPA and optical damage. If thematerial has a large two-photon absorption cross-sec-tion, the loss of light as a function of intensity due toTPA will be larger than the intensity change due tothe phase shift. Mizrahi et al. defined a two-photon

w xfigure of merit 7 , which is proportional to the ratioof the imaginary part of the third-order susceptibility

Fig. 4. Measured phase shifts as a percentage of p.) ISQ switch-ing efficiency is 50%rW.

to the real part. The constant of proportionalitydepends on the type of device. We have found thatthe HSQ dye, while not showing the largest switch-

Žing efficiency, has the smallest figure of merit -1.in the worst case and is therefore most appropriate

for devices. We determined the imaginary part of thesusceptibility using the analysis described in a previ-

w xous paper 6 . Unfortunately, the uncertainty of thismeasurement is about 50%, so the imaginary valuesare not reported here.

The second issue is optical damage. We havefound a very broad range of damage thresholds offiber materials. It appears that the damage thresholddepends on the processing history of the fiber, so itmost likely depends on extrinsic properties such asmaterials purity and defects. For example, we havefound optical damage in the range from 0.5 GWrcm2

to 4 GWrcm2. One fiber endured 4 GWrcm2 forseveral weeks without any noticeable damage. Thisis in contrast to some fibers that damage at 0.5GWrcm2 after a couple laser shots. The best materi-als can thus tolerate the intensity levels that arerequired for all-optical switching.

2.3. Dual-core fiber switching

Our laboratories have succeeded in making dual-w xcore single-mode fiber with nonlinear dyes 8 , simi-

w xlar in geometry to dual-core glass fiber 9–11 . Whenthe core separation is on the order of its diameter,energy flows back and forth between the cores as thelight propagates down the fiber. Fig. 5 shows anend-view image of the dual-core fiber with ls1.064mm light propagating down the guide. The bottomplot shows the intensity profile through the coresalong the line drawn on the picture. The cores aremade of 0.3 wt.% ISQ dye in PMMA and areseparated by about one diameter. The coupling length— defined as the propagation distance over whichall the energy transfers from one core to the other —of these fibers is measured to be about 9.4 mm.

Because the cores are nonlinear, the couplinglength depends on the intensity of light launched intoone of the fiber cores. The nonlinear coupling effectbetween waveguides was previously observed in sil-

w xica fiber 12–14 , and in a dual-core Er-doped fiberw xnear-resonance 15 . We did a series of experiments

in which the light exiting the fiber is imaged over a


Ž .Fig. 5. Image of guiding light in a dual-core fiber top andŽ .intensity profile through its center bottom .

range of input intensities. Each scan is digitized asshown above to determine the intensity of light thatexits each core. Fig. 6 shows a plot of the normal-

Žized measured intensities exiting i.e., light exitingeach core divided by the total intensity that exits

.both cores as a function of the incident intensity.For a linear waveguide, the result should be indepen-dent of the input intensity. Fig. 6 shows that 10% ofthe light couples from one core to the other over thelength of the 35 cm fiber when the incident peakintensity is varied over a few watts. Note that precisevalues of the intensity of light coupling into the fiberare unknown because of the high degree of variabil-ity of coupling efficiency.

ŽFor the geometry of our fiber i.e. length, core.size, core spacing, refractive index profile and the

incident peak power range, we would expect close tofull switching, that is, all the light in one core shouldhave coupled to the other core over the range ofincident powers. There are several difficulties thatcan explain why full switching was not observed.First, because the cores are close together, it ispossible that when light is launched into one core,some of it couples into the second core. If, forexample, 50% of the light were launched into bothcores, no light transfer would be observed betweencores because for any light transferring from core 1to 2 there is an equal amount transferring from 2 to1.

A more likely cause of incomplete energy transferw xis the inhomogeneity of the fiber 8 . Because the

length of the fiber used in these experiments is about37 coupling lengths, random variations in the refrac-tive index, spacing, and core diameter could lead toincoherent multiple scattering between cores, leadingto about an equal amount of light in each core. Thiswould lead to a decrease in the amount of light thattransfers between cores. Assuming that Fig. 6 repre-sents about 1 cycle, the approximate nonlinear re-fractive index is consistent with the phase shiftsgiven in Fig. 4.

Polymer optical fibers thus appear to be a reason-able waveguide approach to making optical switch-ing devices. Before high quality devices can bemade, more work needs to be done in materials-processing studies and characterization.

Fig. 6. Nonlinear intensity coupling in a dual-core fiber.


3. Optical sensors and photomechanical effects

A sensor is a device that converts one form ofmeasurement to another. For example, a thermistorconverts temperature to resistance — which can bemeasured by an electrical circuit. Optical analogs ofsuch devices are finding applications due to theirhigh sensitivity and immunity to electromagneticpick-up. As an example, fiber Bragg gratings arebeing used as strain gauges in diverse environmentsthat include bridges, spacecraft, and marine vesselsw x16 . The wavelength of light reflected from a Bragggrating depends on its period. When a fiber with agrating is stretched, the period changes and thereflected wavelength shifts. The fiber can thereforebe simply interrogated with a broad spectral source.In general, structures that incorporate such sensorsare often referred to as smart structures.

A photomechanical effect, on the other hand, isthe optical version of a motor. For example, if thelength of a material depends on the intensity of lightillumination, it acts as an optical actuator — that is,light induces mechanical motion. In the followingdiscussion, we give an overview of some recentwork and speculate on applications that would bebased on an optical actuator.

3.1. Origins of photomechanical effects

There are many mechanisms that lead to a pho-tomechanical effect, including photothermal heating,electrostriction, molecular reorientation, and elec-tronic excitation. The magnitude and speed of theresponse depends on material properties and samplegeometry. Each response mechanism is briefly de-scribed below.

3.1.1. Photothermal responseWhen light energy is absorbed by a material and

converted to heat, the material’s temperature in-creases, which, through thermal expansion results ina length change. While this mechanism is slowŽ .millisecond time scale , the magnitude of the effectis large.

3.1.2. ElectrostrictionIt is well known that an electric field gradient,

and hence a gradient in the intensity, results in a

force on a dielectric material. In a single-mode poly-mer optical fiber, the light intensity is peaked at the

w xcenter as shown in the other paper in this issue 1 .At the interface between the core and cladding, thereis a large intensity gradient. The net result is aninward force on the core that acts to decrease thediameter and increase the length. Because fibers arethin, they maximize the intensity. Their long length,on the other hand, leverages small strains into arelatively large absolute length change. As an exam-ple, a 10y4 strain in a 1 m long fiber yields a lengthchange of 100 mm. Such length changes are easilydetectable with optical interferometric techniques.

ŽFurthermore, electrostriction should be fast nanose-.cond response-time .

3.1.3. Molecular reorientationWhen a beam of polarized light illuminates a

molecule, its long axis will align along the electricfield polarization of the light beam. If these moleculesare imbedded in a polymer, the molecule–polymerinteraction will stress the polymer and result in adeformation. This deformation can yield a shapechange of the bulk material. Molecular reorientationin liquids and gasses is a well-known and well-studied

w xprocess with sub-picosecond response 17 . Simi-larly, in solids, this mechanism is related to a reori-

w x Žentational phonon which is also fast 18 resonances.with picosecond periods but the amount of deforma-

tion is small.

3.1.4. Electronic mechanismThe electron cloud of a molecule becomes dis-

torted when it is placed in an electric field. Theinteraction potential between molecules in a bulkmaterial depends on the electron clouds of themolecules. In the presence of a light beam, then, thespacing between molecules will be affected. While

Žthe electron-cloud deformation-time is ultrafast fem-.tosecond it takes more time for the nuclei to move.

The material response time would therefore dependon the structure of the solid.

3.2. OÕerÕiew of photomechanical research

The first demonstration of a photomechanical ef-fect in polymer fiber was based on the photothermal


w xmechanism 19,20 . In this work, a mirror is sus-pended from the end of a vertically-hanging pho-tomechanical fiber in a vacuum chamber. The upperend is held in place by a collet. The mirror definesone arm of an interferometer. A fixed mirror outsidethe chamber defines the other arm. Because theposition of the freely-hanging mirror is determinedby the fiber length, the light intensity exiting theinterferometer is a sinusoidal function of the fiber’sposition. As such, the fiber length is related to thelight intensity.

The light leaving the interferometer is fed backinto the hanging fiber at the collet end as a source ofoptical feedback; that is, the length of the fiberdepends on intensity exiting the interferometer; andthe light intensity leaving the interferometer dependson the fiber length, and so on . . . This device wasfound to be multistable and acted as a vibrationstabilizer. It could hold the length of a 30 cm fibersteady to within 3 nm — 1 part in 100,000,000.What is significant is that it combined all four deviceclasses: The hanging mirror and interferometer actedas the sensor, the interference output intensity was

Ž .encoded with the fiber length logic , the informationŽwas steered around the experiment with prisms in-

.formation transmission and the fiber acted as anactuator. This was therefore the first all-optical de-vice demonstration in which no electronics are re-quired.

This demonstration was followed by the design ofa mesoscale version of the photomechanical stabi-

w xlizer 21 . Fig. 7 shows a diagram of this device in a

general characterization set-up. Because the dimen-sions of this device are in the micrometer to millime-ter range, it is called a mesoscale photomechanical

Ž . Žunit MPU . The MPU is a short piece of fiber froma few hundred micrometer in length up to 1 cm in

.length with reflectors defined on each end. Thesereflectors can be made with a variety of methodsincluding optical formation of Bragg gratings, endmetallization, or making a retroreflector on each end.

The MPUs that we have evaluated operate asfollows. The material is a dye-doped polymer, whichthrough the photothermal mechanism, expands whenilluminated by light. The end reflectors define aFabry–Perot interferometer, so the MPU becomes100% transmitting when the spacing between themirrors is an integral multiple of the wavelength.Furthermore, the intensity inside the MPU is directlyrelated to the transmitted intensity. Optical feedbackis automatically built into the unit as follows. Theintensity inside the MPU depends on the length andthe length depends on the intensity. If, for example,the MPU is compressed, the light intensity insideincreases, and, through the photomechanical effect,the length increases. The net result is that the MPUactively resists changes in its length.

Fig. 7 shows a general characterization experi-ment that uses two lasers, each of a different color.The control laser is launched into the MPU after tworeflections from beamsplitters. Some of the light

Ž .passes the first beam splitter BS a1 and hitsdetector 3. Detector 3 thus monitors the control laser

Žintensity. A second laser diode called the probe

Fig. 7. A mesoscale photomechanical unit in a general characterization experiment.


.laser is directly coupled into the MPU after passingŽ .the other beam splitter BS a2 . Some of this beam

Ž .is deflected to the detector a1 . The optical filter’sspectral properties are such that it absorbs the con-trol-laser wavelength but passes the probe laserdiode’s wavelength. Detector 1 thus monitors theprobe laser diode’s output.

The intensity transmitted through the MPU ismonitored by diode 2. A filter at the output can beused to select one of the laser diode wavelengths.Part of the light reflected by the MPU is reflected byBS a2 and transmitted through BS a1 to detector 4,which can be used to monitor one or both beamsŽ .depending on filter choice .

In a typical characterization experiment, the con-trol laser power is adjusted while both the probeintensity transmitted through and reflected from theMPU is monitored. The control laser thus changesthe length of the MPU while the probe laser moni-tors the change. Alternatively, a speaker can be usedto excite vibrations in the MPU so that the effect ofthe photomechanical response on mechanical im-pulse can be studied.

Fig. 8 shows the results of a single laser experi-w xment 21 . The incident laser is ramped up and down

and the transmitted versus the incident intensity isplotted. The observed hysteresis is predicted for afeedback system in which the phase of the light inthe device depends on intensity. There are two possi-ble sources of phase shift; an intensity-dependentlength change and an intensity-dependent refractiveindex. Calculations show that both contribute.

Fig. 9 shows the results of a two-beam experi-ment, in which the transmitted probe intensity is

Fig. 8. Output as a function of input intensity in an MPU.

Ž .Fig. 9. A probe beam points and guide to eye is modulated withŽ .an MPU using a pump beam bottom curve .

measured as a function of time when the controllaser is turned on and off. This is an example ofall-optical modulation in which one beam can con-

w xtrol the intensity of a second beam 22 . Here, the 1cm long MPU is acting as a tunable interferencefilter where the control beam varies the spacingbetween the reflectors.

Fig. 10 shows an expanded view of the area in thedashed box in Fig. 9. The probe data is fit to a risingexponential from the time at which the pump pulse isturned off and a decaying exponential when thepump is turned on. The response time so determinedis 100 ms. Our calculations show that the responsetime depends on MPU size and the thermal conduc-tivity of the MPU and its surroundings. In principleŽ .under ideal conditions , the response time can ap-proach the microsecond range.

Applications of MPUs have already been re-ported. For example, an MPU was attached to a

Fig. 10. Exponential decay of MPU length. Points show data andexponential curves are theory.


Fig. 11. Two interacting MPUs can be used for complexsensingrprocessing.

‘‘drumhead’’ which is made of a mylar sheet that isstretched over the opening of a 1 cm diameter aper-

w xture 23 . When the MPU is powered with light, itactively suppresses vibrations in the drumhead.

3.3. Future applications

It is interesting to speculate on applications ofpolymer optical fiber. Given that polymers have beendemonstrated to be capable of all four device classes,high performance hybrid devices that employ onlyoptics may be envisioned. While even a single com-ponent feedback system has a complex response,when many such components are combined and al-lowed to interact, the result may be an ultra-smartsystem that is capable of performing complex rea-soning.

Consider, for example, a system of two MPUunits as shown in Fig. 11a. The one closer to thelaser diode is transparent and is designed to have nophotomechanical response — it acts as a passivesensor. The absorbing MPU, however, passes lightdepending on both the local stress and the intensityof light incident on the MPU. The MPU’s photome-chanical response will therefore be determined fromthe stresses on each MPU. Furthermore, light re-flected from the active MPU will interact with thetransparent MPU, which in turn reflects the lightback to the active MPU. This added interaction addsa degree of complexity that can be used to ‘‘pro-gram’’ the pair so that the response of the activeMPU is a predetermined function of the geometricalarrangement of MPUs. This programming is achievedby picking the appropriate reflector spacing and dis-tance between MPUs.

Fig. 11b shows a system that behaves similar toFig. 11a except that the devices are wavelength

Fig. 12. A polymer fiber with many MPUs.


Fig. 13. Making a grating reflector through optical writing usingthe interference pattern between two lasers.

addressable. That is, at one color MPU 1 is activeand MPU 2 is passive while at another color, theroles are reversed.

Finally, we can imagine a string of MPUs asdepicted in Fig. 12. In analogy to the dual MPUdevice, these associated networks of devices can bedesigned to be wavelength addressable and eachdevice can be designed to interact with the fullensemble of all other devices in the chain. As such,each device would respond according to the stressesdistributed along the chain of MPUs as well as theoptical state of all the other MPUs. Polymer fiberscould thus be used as the basic thread in makingsmart materials and smart structures.

The fabrication process for producing a string ofMPUs is compatible with the fiber drawing process.One method for making a grating reflector is to crosstwo ultraviolet rays from outside the fiber as shownin Fig. 13. The period of the intensity grating de-pends on the crossing angle between the two beams,so a grating with a specified period — and thereforethe spectral response of its reflectivity — can beburned into the fiber through cis–trans isomeriza-

w xtion or physical surface relief grating formation 24 .A pulsed laser system could be used to burn gratingsjust before the take-up spool in the fiber-drawingapparatus, making it possible to fabricate many MPUsin long lengths of fiber.

The area of smart materials and smart structuresusing light as the power source is a new and unex-plored area in which basic research is being appliedto understanding photomechanical mechanisms forthe purpose of determining structure–property rela-tionships.

4. Conclusion

In conclusion, we have demonstrated that polymeroptical fiber can be used to make all-optical switches,sensors and actuators. Furthermore, we have demon-strated an all-optical positioner and stabilizer thatuses all four device classes, and have speculated onsmart materials and structures. The set of devicesdescribed in this paper, and in our related paper inthis issue on electro-optic fiber and transmission

w xfiber 1 show that polymer fiber offers a palette ofcomponents that can be used to build devices thatcombine the four device classes.

Acknowledgements

We thank the Air Force Office of Scientific Re-search and Army Research Office for generouslysupporting this work.

References

w x1 M.G. Kuzyk, D.W. Garvey, B.K. Canfield, S.R. Vigil, D.J.Ž .Welker, J. Tostenrude, C. Breckon, Chem. Phys. 245 1999

327.w x Ž .2 N.J. Doran, D. Wood, Opt. Lett. 13 1988 56.w x Ž .3 K.J. Blow, N.J. Doran, B.K. Nayer, Opt. Lett. 14 1989 754.w x4 M.N. Islam, E.R. Suderman, R.H. Stolen, W. Pleibel, J.R.

Ž .Simpson, Opt. Lett. 14 1989 811.w x5 M.C. Gabriel, N.H. Whitaker, C.W. Dirk, M.G. Kuzyk, M.

Ž .Thakur, Opt. Lett. 16 1991 1334.w x6 D.W. Garvey, Q. Li, M.G. Kuzyk, C.W. Dirk, Opt. Lett. 21

Ž .1996 104.w x7 V. Mizrahi, K.W. Delong, G.I. Stegeman, M.A. Saifi, M.J.

Ž .Andrejco, Opt. Lett. 14 1989 1140.w x8 S.R. Vigil, Z. Zhou, B.K. Canfield, J. Tostenrude, M.G.

Ž .Kuzyk, J. Opt. Soc. Am. B 15 1998 895.w x Ž .9 B. Diano, G. Gregori, S. Wabnitz, J. Appl. Phys. 58 1985

4512.w x Ž .10 S.M. Jensen, IEEE J. Quant. Electron. 18 1982 1580.w x Ž .11 A. Maier, Sov. J. Quant. Electron. 12 1982 1490.w x12 D.D. Gusovskii, E.M. Dianov, A.A. Maier, V.B. Neustreuev,

E.I. Shklovskii, I.A. Scherbakov, Sov. J. Quant. Electron. 14Ž .1987 1144.

w x13 S.R. Friberg, Y. Silberberg, M.K. Oliver, M.J. Andrejco,Ž .M.A. Saifi, P.W. Smith, Appl. Phys. Lett. 51 1987 1135.

w x14 S.R. Friberg, A.M. Weiber, Y. Silberberg, B.G. Sfez, P.S.Ž .Smith, Opt. Lett. 13 1988 904.

w x15 R.A. Betts, T.T. Jungiarto, J.L. Xue, P.L. Chu, IEEE J.Ž .Quant. Electron. 27 1991 908.


w x Ž .16 E.J. Friebele, Optics & Photonics News 9 1998 33.w x17 J. Etchpare, G. Grillon, J.P. Chambaret, G. Hamoniaux, A.

Ž .Orszag, Opt. Commun. 63 1989 329.w x18 M.G. Kuzyk, C.W. Dirk, Characterization techniques and

tabulations for organic nonlinear optical materials, MarcelDekker, New York, 1998, p. 172.

w x Ž .19 D.J. Welker, M.G. Kuzyk, Appl. Phys. Lett. 64 1994 809.

w x Ž .20 D.J. Welker, M.G. Kuzyk, Nonl. Opt. 15 1996 435.w x Ž .21 D.J. Welker, M.G. Kuzyk, Appl. Phys. Lett. 66 1995 2792.w x Ž .22 D.S. Welker, M.G. Kuzyk, Appl. Phys. Lett. 69 1996 1835.w x Ž .23 D.J. Welker, M.G. Kuzyk, Opt. Lett. 22 1997 417.w x24 X.L. Jiang, L. Li, J. Kumar, D.Y. Kim, S.K. Tripathy, Appl.

Ž .Phys. Lett. 72 1998 2502.